Slashdot Mirror

← Back to Stories (view on slashdot.org)

Inferring Personality From Email Addresses

Posted by CmdrTaco on from the i-really-don't-even-care-for-tacos-that-much dept.

paleshadows writes "Three researchers from the University of Leipzig published an interesting paper titled 'How extroverted is honey.bunny77@hotmail.de? Inferring personality from e-mail addresses' (PDF). From the abstract: 'Email addresses represent the thinnest slice of information that people receive from one another. Using 599 e-mail addresses of young adults, their self-reported personality scores and the personality judgments of 100 independent observers, it was shown that personality impressions based solely on e-mail addresses were consensually shared by observers. Moreover, these impressions contained some degree of validity. This was true for neuroticism, openness, agreeableness, conscientiousness, and narcissism but not for extroversion."'

3 of 309 comments (clear)

  1. Bad example? by Prien715 · · Score: 5, Insightful

    So the article summary starts with:
    How extraverted is honey.bunny77@hotmail.de? Inferring personality from e-mail addresses

    And ends with:
    Moreover, these impressions contained some degree of validity....but not for extraversion

    So the only example in the summary is wrong. And you can tell by reading the summary. Bravo.

    --
    -- Political fascism requires a Fuhrer.
  2. why don't you email him by toby · · Score: 4, Insightful

    and ask.

    --
    you had me at #!
  3. Re:What next? by bob_herrick · · Score: 5, Insightful

    Bollocks. The correlations are weak, at best, and are barely distinguishable from chance. E.g., Consider Table 2. There are 196 entries in the top section of the table. The excess of postive correlations over negative corrlations is a grand total of 8. Assuming 50/50 odds, that excess will happen about 11% of the time just by chance alone. When you factor in the conditional probablity of publishing results (i.e., the argument that if they were any weaker, the data would never have been published), this has to be an extraordinarily weak finding.

    The average correlation (without regard to sign) in the same section of Table 2 is a whoppping 0.067, suggesting an average explanatory power on the order of 0.5%. I suppose such power might have some benefit to someone that sends a lot of e-mails to random addresses like spammers, but for the odinary Joe or Jo, this is not a lot to go on.